Abstract
The problem of tracking human activities of daily living is considered an important subject that crosses a broad spectrum of disciplines. In this paper, we attempt to combine diverse and uncertain measurements arising from observing the daily routines of inhabitants in a closed environment such as smart homes. Specifically, we design a system that adapts itself by learning essentially the inhabitant lifestyle. The problem connotes the exceptionality, doubt, and lack of regularity that characterize generally the inhabitant activity. To tackle this difficulty, we provide a Bayesian model to perform a dynamic elucidation of the cause-effect relationships between some events and certain activities we need to legitimate at a given moment. The model includes an indication about the mean time spent to execute each activity which contributes to enhance the decision rule during the system recognition.
